Fibronectin forms the most extensible biological fibers displaying switchable force-exposed cryptic binding sites

Force to Induce Assembly of Fn into Fibers.

The force measurements were performed using a force sensing lever arm fabricated by silicon MEMS technology (FT-G30; FemtoTools) (see SI Text), as μN were needed to pull the fibers that are out of range for magnetic tweezers and laser traps (20, 21). This force sensor was inserted into a droplet of Fn in PBS (Fig. 1A), force was measured during withdrawal of the tip (Fig. 1B), and the emerging fiber was concurrently imaged with confocal laser scanning microscopy. The plot of force versus time in Fig. 1B shows several regimes. Initially, the tip deformed the air-water surface of the droplet causing a linear increase in force. Once the tip separated from the droplet surface (Fig. 1A, image II), the force remained constant (Fig. 1A, image III), followed by another linear increase (Fig. 1A, image IV), whereas the control, pulling out of the plain air-water droplet, showed forces <1 μN, indicating that the forces mainly originate from the transfer of a Fn layer at the air-water droplet surface into the emerging fiber.

Extensibility of Fn Fibers Until Breakage Greatly Exceeds Earlier Observations.

To quantify the extensions at which the fibers break, we manually pulled Fn fibers from droplets and deposited them across microfabricated trenches on a stretchable PDMS sheet (Fig. 2 A–C). Immediately after deposition and before all subsequent measurements, the Fn fibers were rehydrated in physiological PBS buffer. Since the fibers were held under tension until their deposition onto PDMS, it first had to be determined how much these freely suspended fibers crossing the trenches can be relaxed. We therefore prestrained the PDMS sheet, deposited 38 different fibers across the trenches, and then slowly reduced the mechanical strain of the PDMS sheet until the Fn fibers began to sag (see also SI Text). The fiber strain axis for Fn fibers was normalized to the point at which 50% of these fibers were no longer straight and mechanical tension was absent (0% strain thereby defining L_o), which occurred when fibers were relaxed by 2.4-fold. The fiber length was determined by measuring the channel width. Percent fiber strain is thus defined as 100*(L-L_o)/L_o, where L is the fiber length. In a second set of experiments, we deposited Fn fibers on unstretched PDMS sheets and determined the percentage of fibers that were still intact and not broken by stretching and analyzing 296 fibers (Fig. 2D). While the very first fibers broke at a 2-fold extension (100% strain), some fibers remained intact even after an 8.6-fold extension (760% strain). Half of the fibers were broken at 8.4-fold extension (740% strain). To exclude that the extension originates in part from irreversible material slippage from the fibers which are immobilized on the PDMS ridges toward the highly strained fibers spanning the trenches, we photobleached the fluorophores within small sections of Fn fibers at the sides of the PDMS trenches and tracked the position and shape of the photobleached spots while straining the system (Fig. 2C). These data confirmed that neither slippage between the fiber and the PDMS plateau surface nor differential shearing within the Fn fiber occurred since fluorescence did not penetrate the photobleached sections of the fiber (see also Fig. S1). Once Fn fiber strain exceeded 850%, which corresponds to a 260% relative strain of the underlying PDMS sheet, the PDMS sheets began cracking and tearing, and we therefore did not analyze further extensions. The extensibility of Fn fibers is thus considerably higher than estimated from previous experiments, where data from a small number of ECM fibers that were broken in a living cell culture showed contractions to about one-third to one-fourth of the starting length (11, 12). It was not determined if these fibers had been stretched maximally before release.

The Stress Versus Strain Curves of Single Fn Fibers Are Highly Nonlinear.

To measure the stress-strain curves of single fibers, the MEMS force sensor tip was first brought into contact with the Fn fiber and then displaced along the trenches within the upper plane of the PDMS sheet (Fig. 1 B, C, E, and F). The tensile force applied to strain the fiber, calculated stress (force per unit cross-sectional area of the Fn fiber), and the instantaneous Young's modulus (slope of the stress versus strain curve; Fig. S2) are given as functions of the optically measured fiber extension for one representative fiber in Fig. 1 D and G (six additional fibers are shown in Fig. S3). To determine the cross-sectional area, the diameters of the fibers were measured optically at zero strain. When stretching the fibers, we assumed that the fibers maintained constant volume and a uniform, circular diameter (SI Text and Fig. S4). The Young's modulus of single Fn fibers (Fig. 1G) is not constant as expected for linearly elastic materials that can be approximated by a Hookian spring model. Instead, the stress-strain curves are highly nonlinear, being soft (compliant) first, and turning rigid at high extensions. The Young's modulus exhibited a small increase in the low strain regime, while increasing considerably once the fiber extension exceeded about 150% strain. Note that the Young's modulus of single fibers changes orders of magnitude, e.g., from less than hundred kPa to several MPa going from relaxed to a highly stretched fiber.

While the pulled fibers were 3.5 ± 0.2 μm in diameter, smaller forces would be needed to stretch ECM fibers, where the majority measure a few to hundreds of nanometers and rarely reach micrometers in diameter. Cells typically apply nN forces to cell adhesion sites (22), and cell-generated forces are sufficient to stretch and unfold the much thinner ECM nanofibers (16, 18, 19). Importantly, probing the strain of stretched proteins is physiologically more significant than the force, since the strain determines the corresponding fiber rigidity, and whether a switch in the structure-function relation has occurred in response to tensile force.

FRET Indicates that the Extraordinary Fn Fiber Extensibility Is Due to Force-Induced Structural Unfolding.

Previous studies using FRET indicated that fully relaxed fibers have some residual quaternary structure and that fiber stretching progressively unfolds Fn modules (15, 18, 19). The free cysteines in FnIII₇ and FnIII₁₅ were labeled with acceptors, while lysines were randomly labeled with donors. To determine how the nonlinear increase of the stress-strain curve and the Young's modulus correlate with the unfolding of FnIII modules, Fn fibers containing <5% of Fn-DA were deposited across PDMS trenches. After relaxing them, they were stretched in a stepwise manner, where the peak intensities of the acceptor and donor fluorophores were determined (I_A/I_D; see SI Text). In Fig. 2E, we see that the percentage of structurally perturbed Fn as probed by FRET is small below strains of 100%, but then increases steeply (>100% strain; see Fig. S5). Note that the acceptors are site-specifically bound to the cysteines on FnIII₇ and FnIII₁₅ and thus only probe structural perturbations in their neighborhood, i.e., at donor-acceptor distances below 10 nm. Starting from relaxed fibers, tracking the same individual fibers is only possible to about 300% strain (limited by the maximal extension of the underlying PDMS substrate).

Stretching Steadily Increases the Exposure of Cryptic Binding Sites on FnIII Modules.

To ask whether Fn fiber stretching can expose cryptic binding sites and how the exposure of cryptic binding sites scales with the corresponding stress-strain curve, we took advantage of the fact that both FnIII₇ and FnIII₁₅ each have a cryptic-free cysteine, which is normally only exposed upon denaturing. For intracellular proteins, cryptic cysteines have been exposed by cell-generated forces that were then reacted with fluorophores (23). We adopted this approach to quantify the number of cryptic cysteines exposed by force by allowing maleimide-conjugated Alexa 488 fluorophores in solution to react with the previously unlabeled fibers [described in (23) with more details in SI Text]. After quantifying the number of cysteines exposed on relaxed fibers, Fn fibers were then stretched stepwise (Fig. 2F), and after each step, the fibers were blocked with albumin for 15 min to prevent unspecific binding and again incubated with Alexa 488 conjugated with maleimide. Control experiments showed that this labeling time was sufficient to saturate the exposed cysteines such that incremental labeling after straining the fibers further originated from newly exposed cysteines. The average cysteine exposure determined by fiber fluorescence intensity as a function of strain indicated that Alexa 488 maleimide binding started to occur already below 150% strain. Above 150% strain, the exposure of buried cysteines increased sharply and steadily with increasing strain.

Several conclusions can be drawn from this set of experiments. Most importantly, the number of stretch-exposed cryptic binding sites increases steadily over the entire range of fiber extensions. FnIII₇ and FnIII₁₅ unfolding begins already in the low strain regime where some cryptic binding sites become exposed. This might be unexpected, since some residual quaternary structure exists, as well within fully relaxed Fn fibers as concluded from earlier FRET studies (19). It indicates that fiber stretching does not first fully eliminate quaternary structure before the onset of module unfolding, and that both processes seem to occur in parallel in the low extension regime. Furthermore, the strain-dependent rate at which these buried cysteines become exposed directly parallels the increase of module unfolding as probed by FRET (Fig. 2E). We previously interpreted the FRET ratios of mechanically unfolded Fn by comparison to Fn that was denatured in solution (15, 19). In agreement, the correlations presented here between stretch-induced cysteine exposure (Fig. 2F), loss of FRET in stretched fibers (Fig. 2E), and loss of FRET based upon chemical denaturing in solution (Fig. 2E and Fig. S5) demonstrate that the FRET ratios relate to loss of secondary structure of the FnIII₇ and FnIII₁₅ modules and their surroundings (also see Fig. S5).

Fiber Contraction and Module Refolding Occurs When Tensile Force Is Released.

Since many materials deform plastically before failure, whereby plastic deformations involve irreversible slippage of molecules with respect to each other, we studied whether the fibers returned to their original length after stress release, and if so, how rapidly. The contour length of fibers was measured in time-lapse microscopy after fibers were released from the MEMS tip. Immediately after release, the fibers had a sinuous appearance and then contracted back to their original length as time progressed (Fig. 3 A–D). For every fiber in the strain regime investigated, the fiber length always returned back to the starting length after a recovery period of a few minutes regardless of the state of extension (Fig. 3E). The contraction kinetics of three individual fibers is shown in Fig. 3E. The optically measured time-resolved contraction of the contour length of single fibers can be fitted by a double exponential. While each fiber has slightly different kinetics of contraction, all of these representative three fibers have a fast and a slow recovery regime (τ₁ = 2.1, 2.0, and 1.7 s, and τ₂ = 21, 38, and 83 s). Before releasing the tensile force, these three initially relaxed fibers were strained to A₁ = 141, 237, and 259%. Despite the differences in initial strains and corresponding time constants, the double exponential fits yield essentially equal values for the three fibers, i.e., A₂ = 15, 15, and 17%. When probing higher strains and allowing fibers to relax not to 0%, but to 140% strain, their optically measured length recovered within less than a second (see Fig. S6).

To determine whether a complete recovery of the mechanical properties occurs as well, the force-strain curves of the same fiber were probed during multiple extensions with a variable waiting period between each mechanical test (Fig. 4). When the waiting time between consecutive pulls was shorter than 1 min, the force-strain curves showed that the second fiber extension required significantly lower stress to reach a given strain relative to the first fiber pull, and this is true when expanding from a relaxed state, as well as from 140–150% upwards (Fig. 4A). Thus, the energy needed to strain the fiber, i.e., the area under the force versus strain curve, was lower on the second pull after waiting for just a short time. Right after the fibers started to contract, they are thus initially more compliant than the original fibers. However, waiting for 1 min or more resulted in a recovery of their initial mechanical properties (Fig. 4B). This is a significant observation, since mechanical unfolding of Fn modules requires that clusters of force-bearing backbone hydrogen bonds are broken (24). A complete recovery of the mechanical strength of the fibers therefore implies that these critical hydrogen bonds can reform. From an integration of the stress-strain curves of the first and second pulls as a function of the waiting time, we estimated the kinetics of mechanical recovery (Fig. 4D). Within experimental error, the kinetics of reestablishing the mechanical strength were similar, no matter if the fibers were strained from a relaxed state or from 140% (Fig. 4 A and B). The mechanical recovery kinetics of 10 fibers, either strained to about 100% from a relaxed state (green) or strained to about 200–300% and allowing to relax to 140% (red), can both be fitted to the function E₂/E₁ = 1 − exp(-t/τ), where E₁ and E₂ are the integrated areas under the first and the second force versus strain curve, respectively. The resulting time constants to recover the mechanical properties of weakly and highly strained fibers are with τ′ = 58 ± 8 s and τ″ = 38 ± 6 s, i.e., on a similar order of magnitude. Importantly, the mechanical recovery kinetics has the same order of magnitude as τ₂, the slow component of the length recovery (Fig. 3E). We thus conclude that the slow length recovery kinetics is dominated by the rate at which the original force-bearing backbone hydrogen bond network, or the secondary structure of Fn modules, is reestablished.

Discussion and Conclusions

A combination of microfabricated surfaces, spectroscopic imaging of labeled protein, and a MEMS capacitive force sensor was used to characterize the mechanical and structural properties of individual, freely suspended Fn fibers. With more than 8-fold extension (700% strain) with respect to the resting length before 50% of the fibers are broken, the Fn fibers are one of the most extensible macromolecular fibers described so far (Fig. 2). The extraordinary fiber extensibility requires the mechanical unfolding of secondary structure (Fig. 2E), and remarkably, release of mechanical stress permits recovery of the fiber's original mechanical properties (Fig. 3). Given the diameter of the fibers at resting length, this corresponds to more than a million Fn molecules that have to refold synchronously in the crowed environment of a fiber because a module can only refold if the lateral neighbor refolds as well. Upon the release of tensile force, the fibers rapidly contract within seconds (Fig. 3). The fast kinetic contribution of fiber contraction, τ₁, is driven by the contraction of unfolded Fn modules into random coil conformations that initially act as entropic springs, whereby the terminal ends are pulled inwards (Figs. 3 and 4). With a slower time constant (i.e., over minutes), the force-bearing backbone hydrogen bond networks are reestablished (τ₂ in Fig. 3E equaling τ′ and τ″ in Fig. 4D), as reflected by the recovery of the original mechanical strength (Fig. 4). Finally, we demonstrate that fiber stretching exposes cryptic binding sites and that the number of exposed cryptic sites increases first slightly and then steeply once the strain exceeds 100% (Fig. 2F). Cryptic molecular recognition sites can thus be switched on by force as demonstrated here for the exposure of cryptic cysteines on FnIII₇ and FnIII₁₅.

Upon Fiber Contraction, Short-Lived Intermediate Molten Globule States Dominate Before the Force-Bearing Hydrogen Bond Networks Are Reestablished.

Insights into the fiber relaxation mechanism can be derived from the finding that the length recovery kinetics show a double exponential decay, with the time constants for the fast (τ₁ = 2.1, 2.0, and 1.7 s) and the slow (τ₂ = 21, 38, and 83 s) recovery components determined for three representative fibers (Fig. 3). The existence of two kinetic regimes is also apparent when probing how fast the mechanical properties can recover once a stretched fiber is released. Full recovery of the original stress-strain curves upon extension, were typically observed after a waiting period of a 1 min or more (Fig. 4D). Since mechanical unfolding of secondary structure occurs during fiber extension, as indicated by both FRET (Fig. 2E) and the exposure of the buried cysteine residues (Fig. 2F), restoring the force-bearing hydrogen bonds that stabilize the β-sandwich motif of Fn modules is needed to fully recover the mechanical properties of the fibers. It is remarkable that this can happen in a contracting fiber where up to a million proteins have to refold in parallel.

One explanation for the double exponential length recovery once the force is released is that the highly stretched and unfolded modules initially behave like entropic springs and rapidly recoil (τ₁) into molten globule states as the end-to-end distances contract. The existence of short-lived random coiled or molten globule states was initially documented in the refolding pathways of denatured proteins and later in those of mechanically unfolded single proteins (25–27). The lifetime (τ₂) of the random coiled intermediate is determined by the rate at which the force-bearing backbone hydrogen bond networks reestablish (27). In solution, most FnIII modules refold within 200 ms to seconds (28–30). Here we show the existence of a molten globule-like intermediate in the refolding pathway of supramolecular assemblies, and importantly, these short-lived random coiled intermediates may have significantly longer lifetimes (i.e., minutes) in the crowded environment of a fiber. While crowded environments typically increase refolding kinetics of individual molecules (31), our data indicate that the synchroneous refolding of a large number of modules needed to contract Fn fibers greatly slow down the kinetics. Furthermore, the amplitude of the slow kinetics of the length recovery, A₂ in Fig. 3E, corresponds to only 15% strain. This may suggest that the average end-to-end distances of individual modules forming a random coil have to first come sufficiently close to initiate the rebuilding of the hydrogen bond network and thus secondary structure. However, even the fibers held under some residual tension can ultimately recover their mechanical strength if held at a prestrain of 140% (Fig. 4, red data points). Even if not resolvable in the optical contour length measurements, this might indicate that hydrogen bond formation can occur at least in the presence of some residual stress acting on the fiber termini. While we have optically probed the length of the fiber spanning microfabricated trenches, thermally driven end-to-end distance fluctuations of individual modules will be present. Local thermal length fluctuation (facilitated by the small slope in the stress-strain curves at small extensions) potentially allow a biased refolding of the fraction of Fn modules that happens to have an end-to-end distance approaching A₂ (i.e., 15% strain). Since most of the Fn modules carry molecular binding sites (7), our data show that these recognition switches can be turned on and off reversibly, even if assembled into dense fibrils. In summary, for highly strained fibers, the major contribution to the length recovery is driven by the entropic contraction of previously unfolded Fn modules, while the kinetics of reestablishing the original mechanical fiber properties are slowed significantly, dictated by the rate at which the backbone hydrogen bond networks reform.

Physiological Implications.

How did the mechanical properties of Fn fibers adopt to their function? Similar to other biological fibers such as silk and fibrin (1), Fn fibers reach a Young's modulus of a few MPa at breakage (Fig. 1G and Fig. S5). Thus, Fn fibers are neither the strongest nor the toughest biological fibers known, but they can be extended at much lower stresses than silk fibers that are designed to withstand external impacts (32). With only less than 50% of the fibers broken at 8-fold extension, the extensibility of Fn fibers exceeds that of other reported biological protein fibers. From FRET studies in cell culture, we know that cell-generated forces are sufficient to stretch ECM fibers through the full strain regimes (15, 18, 19). Consequently, Fn fibers are a material designed to maximize how much they can be reversibly stretched by cell-generated forces. This is equivalent to maximizing the mechanical energy that can be stored. Furthermore, our data now quantify how fiber extension up-regulates the rigidity of individual Fn fibers. This is highly significant, since cells pull locally via their adhesion sites just on individual fibers and sense and respond to their rigidity. The rigidity of a single fiber could be significantly different from the average rigidity typically reported for macroscopic matrices or substrates. Previous studies in cell culture have shown that the Fn fibers assembled by fibroblasts on soft substrates (7.5 kPa) are barely unfolded, while they are highly unfolded on rigid substrates (33 kPa) (33). This would imply that the Young's modulus of individual fibers stretched by cells sitting on soft substrates are a few kPa, while some fibers reach up to a few MPa in the early matrix of cells sitting on rigid substrates.

Parallel to the increase in fiber rigidity, fiber stretching further gradually switches the biochemistry displayed by the fibers (Fig. 2 E and F). Fn contains a large number of molecular recognitions sites to proteins and cells. Some of these binding sites are exposed on the surface of the folded modules, which might thus be deactivated upon unfolding, while others are buried in the folded modules and can only be exposed upon unfolding (7, 18). Although probing for exposure of cysteines in FnIII₇ and FnIII₁₅ (Fig. 2) here, a similar dependence of cryptic epitope exposure on strain is expected for the stretch-induced unfolding of the other FnIII modules as well. FnIII modules vary about 3-fold in their mechanical stability due to amino acid sequence variations, with FnIII₇ being one of the most mechanically stable FnIII modules (24, 29).

Once cells release a stretched Fn fiber, is it of physiological significance that the elastic energy stored in a stretched fibers can be released by displacing matrix fibers to which they are physically connected? Our data indicate that Fn fiber contraction in the fast regime generates significant tensile forces since the unfolded modules contract like entropic springs. Significantly, it is the fast entropic contraction that contributes most to pulling the opposing fibers ends toward each other. In contrast, the recovery of secondary structure contributes far less to the length recovery and might be slowed or inhibited when the fibers are stretched. Intrinsic to entropic springs is that mechanical energy can be stored by stretching and unfolding of the many Fn modules. This could play a so far not recognized role in early development (34), matrix biology, and wound healing.

Microfabrication of Trenches in PDMS Sheets and Fn Fiber Mechanical Testing.

Standard photolithography was used to structure the PDMS sheets (35) before the deposition of Fn fiber pulled out of droplets of soluble Fn (19, 36) (Fig. 2 and SI Text), rehydrated, and stretched or relaxed (see Fig. 2E and Fig. S5). For the stress calculation in Fig. 1D, the initial diameter was taken, assuming a constant volume of the fiber throughout the measurement (Fig. S2). In addition to Fig. 1G, the mechanical characteristics for six more fibers can be found in Fig. S3.

Photobleached Fiber Segments.

Additional to the bleached fiber segments in Fig. 2C, two additional experiments of photobleached fibers before and after strain application are given (Fig. S1).

MEMS Force Sensor.

The working principle of the capacitance based MEMS sensor can be found in (37) and the SI Text. The linearity for the force measurements were tested plotting optical measured versus calculated (from spring constant) tip displacement (see Fig. S4).

Stretch-Induced Exposure of Cryptic Cysteines.

Fn fibers were deposited over PDMS trenches and incubated at different strains with Alexa 488 maleimide to covalently label the exposed cysteines as function of the fiber strain (Fig. 2F and standard protocols in SI Text) (23).

Length Recovery Kinetics.

The recovery kinetics for a prestrained Fn fiber (140%) that was strained to 280% is shown in Fig. S6.

Confocal Microscopy and Image Processing.

All confocal images were acquired with an Olympus FV1000 confocal microscope with a water immersion 0.9 NA 40× objective. Data processing was done using MatLab custom scripts (MathWorks) (see Fig. 1 and SI Text).

Fn FRET Labeling and Denaturation in Solution.

Published protocols were used to label Fn with donors and acceptors, as well as to chemically denature Fn in solution for FRET calibrations (19). Low percentage of Fn-DA were diluted with unlabeled Fn to prevent intermolecular FRET, as previously described (19) (see Fig. 2E, Fig. S5, and SI Text for further information).